Learning AI: Taking off Your Training Wheels

Recently, I wrote an article about resources you can use to learn AI for free. At the time, I commented that once you get started, you will be coding and may need some support. As a developer myself, I know that you often need other eyes on a problem to gain insight into a sticking point in your understanding.

This article lists a few online communities you might join to augment your learning from reading and videos as you start developing code. As always with online communities, check out the guidance for best practice and remember to “be nice” in your interactions!

AI Communities Online

Of the resources I listed in my original article, the MOOCs — such as those from Coursera and Udacity — come with communities to support you, including access to teaching assistants within some.

Of course, there is always Stack Overflow, and related stack exchanges such as Cross Validated for Machine Learning. But where else can you go to get help when your data pipeline is blocked or you need advice on which algorithm to use or how to tweak your model to improve its accuracy?

Reddit seems to be a popular choice. The Learn Machine Learning subreddit is a great place to start, and there are a number of other subreddits for specific areas of AI, as well as https://www.reddit.com/r/artificial/, which is the subreddit for general discussions about AI.

The Hacker News for data science and beyond is DataTau, which is a good place for sharing data science news, cool stuff you’ve made and tips or tricks such as popular interview questions and cheatsheets.

Over on Github, a user named BAILOOL has a repo with a set of links to country-based AI communities, which mostly seem to be on Facebook.

IRL Communities

A Twitter user brought my attention to an AI meetup group in the Netherlands, which looks really useful and hosts presentations, workshops, hackathons, and co-working sessions. I’m sure there are similar opportunities in other countries if you dig around—if you find any, please share them with us in the comments below. And if not, you could always start your own!

Need Some Ideas?

Having found a community, and decided to get into coding your first few AI projects, you’ll need some data and ideas. Head over to Kaggle, which is another great community resource for data scientists and those working with machine learning. Now part of Google, the site hosts a public data platform where you can find some interesting data sets; they also hold competitions for Machine Learning projects based on the data. (I mentioned Kaggle in my original article about learning AI, because they have some free learning materials which are concise yet comprehensive, covering Machine Learning and Deep Learning).

There are always great ideas for projects using AI, and you can even have some fun. One of my favourite sites is AI Weirdness, by Janelle Shane, who trains neural networks with fun datasets to output some interesting results, such as:

• Metal band names created by a neural net. Is anyone a fan of Sun Damage Omen yet?

• Snake names —  beware the Ned-scaled tree viper

• Heirloom apple types — excuse me if I pass on trying a Failing Puster

• Finally, my all-time favorite is new paint colors invented by Neural Network. Here's a paint chart with some hilarious, AI-generated, paint names.

Shane shares how she works with neural nets if you want to try generating a set of nonsense from a dataset. I’m thinking about using AI to run off a name for a memorable fragrance and a script for its advertisement, based on the terrible voice overs found on recent adverts. You’ll probably find me asking questions over on Reddit.

Thanks for reading and, as ever, please leave your comments below the line, particularly if you have any suggestions for welcoming communities for AI learners. In the meantime, I'm just off to paint my kitchen Stoner Blue...